Target tracking based on multi-color joint probability statistics model
Hongguang Li,Ding Wenrui,Liu Chunhui,Zheng Junling
DOI: https://doi.org/10.1109/CISP.2011.6100008
2011-01-01
Abstract:For real-time target tracking under complex scene, a target tracking algorithm based on multi-color joint probability analysis model was presented. The algorithm adopted color histogram to represent the target statistical characteristic with Camshift principle, and carried out exploratory research in such aspects as multichannel joint color features statistics, projection map area weighted processing, the tracking window size and position calculating, algorithm processing mechanism of circulation. It used red (R), green (G), blue (B), hue(H), luminance (Y) channel color as the target observed characteristics, and designed the calculation method based on the probability statistic to distinguish any color target from complex scene. It also established the calculation method for tracking window size and position which adapted the multi-color model. Using weighting projection map area method, the background interference around the target potential area was eliminated. At last, more reasonable convergence judgment and the algorithm cycle rules were put forward. After the experimental certification, the real-time performance and detection ratio present a good result.